Quality Metrics for Map Databases in Vehicle Routing Applications

By Bert Jakubs

John Pellizzon owns a fleet of trucks related to his trucking business in California."It's not easy finding a customer's house.So when customers call in, our booker looks up their addresses on one of the Internet web sites.Problem is, it's almost as likely to be wrong as it is to be right -- it's frustrating.But having something to go on is better than nothing.">

Pellizzon's experience proves what the industry already knows: your mapping application is only as good as the map database that supports it. When choosing a map database product, it is important to consider the quality of that database.

With so many routing, location based services (LBS), and Global Positioning System (GPS) applications available today, more and more map application developers insist on quality as a fundamental feature of their database products because their users know the difference. However, some users and many developers often think of quality in terms of accuracy and, more specifically, only in terms of positional accuracy of the feature geometry (i.e.how closely the database's representation of geometric features matches ground truth). While positional accuracy is critical, quality considerations should not be limited by it. A good map database designed to support today's routing, LBS and GPS applications should ensure a continual and reliable level of quality that will satisfy certain application categories.

Basic functional requirements also known as application categories include:

  • Geocoding - The map database enables the assignment of latitude/longitude coordinates to street addresses.Pellizzon's booker being able to find customer addresses is an example.
  • Reverse geocoding - The map database enables the retrieval of the street address given the latitude/longitude coordinates.Looking up where his trucks are via GPS latitude/longitude signals gives Pellizzon more control over resources.
  • Routing - Supports the generation of routing directions and real-time route guidance.Pellizzon wants his plumbers making house calls, not getting lost between them.
  • Positioning - Coordinates in the map databases must be sufficiently accurate to support GPS-based applications unambiguously.GPS signals need to match the map accurately so Pellizzon can tell exactly which street his plumber is actually on.
  • Map display - Databases should contain area features such as water and parks that provide context for the road network, as well as enabling an attractive and easy-to-interpret appearance.Pellizzon wants routes that are easily identified; his drivers need to concentrate on their driving.
  • Support of Dynamic Data - Databases should integrate un-translated dynamic data into the database as detailed map elements for accurate functionality. Knowing there's an accident at Interstates 10 and 605 helps a little; knowing it's on the 10 eastbound at the 605 makes a big difference.
Each one of the above application categories has a quality aspect.In order to provide metrics to evaluate and compare quality, the map databases can be viewed as possessing certain critical properties consisting of one or more attributes. Properties form measurable characteristics of map databases that can have profound impact on application performance. Although a map may have many other properties, the road network geometry, feature names, addresses, and routing are the fundamental properties relevant to product applications. The quality metrics of these properties include both completeness and accuracy components.
  • Completeness of a property measures what fraction of real-world items are actually in the database.
  • Accuracy of a property measures what fraction of items in the database is accurately represented.
The measurements are taken by comparing the map database properties against ground-truth or a reliable ground-truth surrogate.The best database providers strive for virtual 100% completeness and accuracy for every map property.
Measurements can be taken for many properties and can be tailored to the specific content of a vendor's map database.With the understanding that quality implies both completeness and accuracy, the following properties in addition to positional accuracy are the most important to consider for routing, LBS and GPS applications.
  • Highway Network Quality
  • Surface Street Network Quality
  • Highway Name Quality
  • Surface Street Name Quality
  • Address Quality
  • Routing Quality
Measurements are taken through a procedure in which the map properties are sampled and then compared to ground-truth.In many cases, ground-truth surrogates that have been validated for completeness and accuracy can be substituted.For example, to measure Surface Street Quality, a current aerial photo might suffice as a ground-truth surrogate.Some properties might require an actual field survey due to unreliable or incomplete availability of ground-truth surrogates.An example is Maneuver or Routing Quality measurement due to the unreliable and incomplete availability of sources for maneuver restrictions from many local and county governments. Therefore, to measure Routing Quality, a statistically valid sample of intersections could be taken and then field inspected to determine the fraction of intersections that have the correct set of maneuvers and guidance features that lead to correct routing.

As the routing, LBS and GPS application market continues to expand and create ever-increasing demands for robust and reliable performance, the demands on map database quality will increase as well. Competition, advancements in technology, and the increasing demand for dynamic content are changing the landscape of mapping products and necessitate accurate and complete map data. Whether your applications support businesses conducting geographic analysis, such as Pellizzon's plumbing business, or helping consumers navigate from start to destination, the quality of application relies on the quality of your map database.To thoroughly understand the quality of your map database, it is important to understand its properties, the functions it needs to support, and how map database quality is approached, executed, and communicated.

Published Tuesday, February 11th, 2003

Written by Bert Jakubs

If you liked this article subscribe to our bimonthly newsletter...stay informed on the latest geospatial technology

Sign up

© 2017 Directions Media. All Rights Reserved.